The Diabetic retinopathy is an eye condition that occurs in the majority of patients diabetes mellitus . It may cause damage to the retina and other eye structures, compromising visual skills. Are affected about 415 million people around the world and the most effective way to combat it is represented by the early diagnosis of the problem.

The examination need to locate it is made from ‘ analysis eye back . An investigation that those at risk should lead every year as a preventative practices. In many, however, especially those who live in areas where access to specialist care is not available to everyone, they are not able and end up realizing it until it is too late. A project fielded by Google in collaboration with team of doctors from India and the United States, aims to exploit the potential of machine learning to offer an effective solution. The results obtained so far, after years of research, have been published since Journal of the American Medical Association .

Examples of photographs of retinal analyzed for this type of pathology: left a healthy eye, the right one that presents lesions

The group of Mountain View has formed a database consisting of approximately 128,000 images used to teach an algorithm to recognize the signs of diabetic retinopathy. Once the appraisal stage, the IA has been subjected to a test: the performance, the ability to recognize the disease, have proved equal to those of a team of professional ophthalmologists.


This is only the first step in a broader process, which in the future will bring a system developed by DeepMind to analyze 3D images of the entire eye and its various layers, not only two-dimensional images as it happens today, thus further increasing the accuracy and the predictive effectiveness of the technology. An artificial intelligence of this kind may one day be used as a support tool for specialists or to provide a first quick diagnosis to patients who have no way to access quickly cures.